As part of a class assignment, I spent 70 minutes on Amazon’s Mechanical Turk as a worker yesterday. I was determined to make it a non-tedious experience, and to learn something about the “future of work.” The idea of a distributed market of isolated tasks, available to millions online initially seemed promising.

Perhaps the mechanical turk model points to the future of work, but it certainly isn’t the future of work.

Mechanical turk is a online marketplace where you can prescribe tasks for people to perform in a web browser for a couple cents to a dollar. If you want someone to perform some sort of data entry work or mechanical research task, mechanical turk might be a good fit, provided that you can describe your task well.

I knew of two rather interesting art projects that made use of Mechanical Turk.  The sheep market project and the ten thousand cents project were both interesting examples of people using mechanical turk creatively. I had some hope that the online market experience would be interesting.

Early reviews were not positive though. Several classmates spoke of mind numbingly boring experience. In an attempt to find something interesting to do, I searched “Art.”

Art apparently corresponded with writing, because all the Human Intelligence Tasks (abbreviated HIT) results were writing tasks which asked me to write a review of electronics or some other consumer products. Essentially asking me to help game Google. I passed on those.

I tried to search for “science” tasks. The HITs that I got were mostly psychology surveys from varies universities. I did a couple of these. I felt more fulfilled by the thought that I was contributing to someone’s psychology research than by the paltry couple quarters that I was getting.

Finally, in a desperate search of a creative project, I tried searching for “draw” - which brought me to a project where I was to draw a bounding box around an object. These kinds of project are usually associated with machine learning. The researchers were likely gathering a data set to train their visual learning algorithms. I like the idea of teaching an algorithm how to recognize objects, so I did these a couple times.

I didn’t find anything more interesting than that though, and I certainly didn’t see any reason why I would return to mechanical turk again.  For my hour, I earned maybe a dollar.

This is a shame. There’s an interesting opportunity to create crowd-sourced art and science projects, but because mechanical turk’s interface is so awful, and so filled with mindless tasks, that the opportunity is lost.

Perhaps there’s an opportunity for a website where people can spend their spare time doing mechanical turk style tasks that are guaranteed to be artistically or scientifically interesting?